Get tomorrow's brief in your inbox
Today: OpenAI split ChatGPT Work into two products (cloud and local), adding code execution with internet access, persistent storage, and sub-agent automation for $20/month subscribers. Caterpillar is applying decades of mining automation experience to deploy AI field tools and invest $100M to train 118,000 employees in AI and robotics.
OpenAI quietly launched two distinct products under the ChatGPT Work brand, and the cloud version is a significant upgrade for business users willing to navigate the confusion. ChatGPT Work Cloud ($20/month minimum) now includes code execution with full internet access, a persistent filesystem shared across sessions, headless browser automation, and the ability to spawn sub-agent sessions with different models. If you've been frustrated by ChatGPT's inability to install Python packages or scrape live data, this fixes it. The code interpreter can now pull from APIs, clone GitHub repos, and interact with external services far beyond Claude's limited domain allowlist.
The catch: OpenAI's documentation is terrible, and the product naming is deliberately confusing. "Work Cloud" runs at chatgpt.com. "Work Local" is the desktop app (formerly Codex) that accesses local files. Both are called "ChatGPT Work." Free users and $8/month Go subscribers are locked out entirely. Model selection differs between Work and Chat (Work offers Luna/Terra/Sol at multiple reasoning levels; Chat offers a different tier structure). Sessions bill against your Codex allowance, not your Chat allowance, so heavy Work users may burn through credits faster than expected.
For small business owners, this matters if you're automating data collection, analysis, or reporting workflows. The persistent filesystem means you can build multi-session projects without re-uploading context. Internet-enabled code execution means you can pull live pricing data, scrape competitor sites (ethically), or automate API-driven tasks without leaving ChatGPT. The tradeoff is complexity: you need to understand which "Work" you're using and when to pick Luna vs. Terra vs. Sol. If your workflows involve repetitive data tasks or multi-step automation, start experimenting with Work Cloud. If you just need answers, stick with Chat.
ChatGPT Work pricing clarification: The $20/month ChatGPT subscription now includes both Work flavors. Free users get nothing. Go users ($8/month) are also excluded. If you're already paying for ChatGPT, you have access. If you're on the free tier and need these features, it's time to upgrade or switch to a competitor like Claude or Perplexity that offers code execution on cheaper plans.
Caterpillar's AI deployment playbook: The industrial giant spent decades automating mining (autonomous haul trucks, drills, loaders) and is now applying those lessons to deploy AI across field operations and enterprise workflows. Their Cat AI Assistant lets technicians use voice commands to pull repair procedures, troubleshoot machines, and identify parts before starting a repair. It runs on 16 petabytes of proprietary data from 1.6 million connected assets. The company is also using AI agents to modernize legacy code, generate and test software, and identify defects earlier.
The insight for SMBs: Caterpillar treats AI deployment as a people problem, not just a technology problem. They're investing $100 million over five years to train 118,000 employees in AI, autonomy, and robotics. Their CTO says the hard part isn't building the tech, it's "incorporating that technology into the customer jobsite and into the workflows." If you're deploying AI tools and hitting adoption resistance, you're not alone. Budget for training, involve experienced operators in building the system, and rethink how people work alongside the technology. Don't just buy the tool and expect magic.
Business impact: Caterpillar's power-generation revenue spiked 72% to $3.10 billion in Q2, driven by data center demand for AI infrastructure. The AI boom is creating indirect opportunities in power generation, cooling, construction, and logistics. If your business serves data centers or supports infrastructure buildout, this trend has legs.
Meta is testing robots in data centers to handle cabling, part reseating, and maintenance tasks. They're piloting dual-armed Watney robots in Altoona and ABB robots with scissor-lift risers in Ohio. Early results: robots are slower than humans, need frequent recharging, and can't handle intense cabling work (like Nvidia GB300 supercomputers). But the direction is clear. If you run physical infrastructure (warehouses, server rooms, manufacturing), watch this space. The robotics industry is moving from demos to real deployments, and the cost curve is dropping fast thanks to Chinese manufacturers shipping 22,000 humanoid units in H1 2026.
Actionable takeaway: Don't redesign your operations for robots yet, but start tracking which repetitive physical tasks could be automated in 3-5 years. The ROI case will improve as robot costs drop and battery life extends.
No major pricing changes today, but the ChatGPT Work confusion underscores a broader trend: AI companies are segmenting access by subscription tier more aggressively. OpenAI now has free, $8/month Go, $20/month, and $100/month Pro tiers with different model access and feature gates. If you're building workflows around a specific AI tool, check the tier requirements before committing. Free tiers are shrinking.
U.S. tariffs on Chinese drones and robots take effect in September, with component tariffs following in 2027. China dominates global humanoid robot manufacturing (86% of shipments in H1 2026 came from five Chinese companies). The U.S. restrictions may protect domestic markets but won't address China's manufacturing scale and cost advantages. Chinese robotics makers are targeting Europe, Southeast Asia, Latin America, and the Middle East instead.
What this means for SMBs: If you're considering buying drones or humanoid robots for warehouse automation, logistics, or inspection tasks, expect prices on Chinese hardware to rise. U.S. and European alternatives will remain more expensive. The market is fragmenting, with Chinese companies expanding globally while U.S. manufacturers compete in security-sensitive markets. If cost matters more than data security, buy before tariffs hit. If you're in a regulated industry (defense, critical infrastructure), budget for higher-cost domestic alternatives.
Texas Governor Abbott froze state spending on Flock AI surveillance cameras after a Texas Tribune investigation revealed the state spent over $30 million on them, funded by a $1 fee on insurance policies. Flock cameras have drawn bipartisan backlash over privacy concerns, data-sharing practices, officer misuse (at least six incidents in Texas), and unsecured video feeds. Multiple Texas cities have canceled contracts.
If you sell to government or operate in public-facing surveillance markets, this is a warning sign. AI surveillance is facing regulatory and public backlash. Privacy controls, transparency, and clear data governance are no longer optional.
Before automating a workflow, document it. Inc.com's automation piece emphasizes the most overlooked skill: understanding your work before you automate it. Most companies skip this step. If you're planning to use AI for customer support, data entry, or reporting, spend a week mapping the actual process (inputs, outputs, decision points, exceptions). You'll catch inefficiencies and edge cases that would break an automated system. Bonus: the documentation makes it easier to train AI tools or hand off tasks to team members.
Two patterns today: AI tools are getting more powerful but more confusing (ChatGPT Work's split product launch is a mess), and real-world AI deployment is hitting people and process problems, not technology limits. Caterpillar's $100M training investment and Meta's slow robot rollout both point to the same lesson: the hard part isn't building AI, it's integrating it into existing workflows and training people to use it. If you're deploying AI this year, budget more time and money for change management than you think you need. Watch for more robotics and automation stories as tariffs reshape the market and Chinese manufacturers push into non-U.S. markets.